xref: /qemu/fsdev/virtfs-proxy-helper.c (revision de6cd759)
1 /*
2  * Helper for QEMU Proxy FS Driver
3  * Copyright IBM, Corp. 2011
4  *
5  * Authors:
6  * M. Mohan Kumar <mohan@in.ibm.com>
7  *
8  * This work is licensed under the terms of the GNU GPL, version 2. See
9  * the COPYING file in the top-level directory.
10  */
11 
12 #include "qemu/osdep.h"
13 #include <glib/gstdio.h>
14 #include <sys/resource.h>
15 #include <getopt.h>
16 #include <syslog.h>
17 #include <sys/fsuid.h>
18 #include <sys/vfs.h>
19 #include <sys/ioctl.h>
20 #include <linux/fs.h>
21 #ifdef CONFIG_LINUX_MAGIC_H
22 #include <linux/magic.h>
23 #endif
24 #include <cap-ng.h>
25 #include "qemu/sockets.h"
26 #include "qemu/xattr.h"
27 #include "9p-iov-marshal.h"
28 #include "hw/9pfs/9p-proxy.h"
29 #include "hw/9pfs/9p-util.h"
30 #include "fsdev/9p-iov-marshal.h"
31 
32 #define PROGNAME "virtfs-proxy-helper"
33 
34 #ifndef XFS_SUPER_MAGIC
35 #define XFS_SUPER_MAGIC  0x58465342
36 #endif
37 #ifndef EXT2_SUPER_MAGIC
38 #define EXT2_SUPER_MAGIC 0xEF53
39 #endif
40 #ifndef REISERFS_SUPER_MAGIC
41 #define REISERFS_SUPER_MAGIC 0x52654973
42 #endif
43 #ifndef BTRFS_SUPER_MAGIC
44 #define BTRFS_SUPER_MAGIC 0x9123683E
45 #endif
46 
47 static const struct option helper_opts[] = {
48     {"fd", required_argument, NULL, 'f'},
49     {"path", required_argument, NULL, 'p'},
50     {"nodaemon", no_argument, NULL, 'n'},
51     {"socket", required_argument, NULL, 's'},
52     {"uid", required_argument, NULL, 'u'},
53     {"gid", required_argument, NULL, 'g'},
54     {},
55 };
56 
57 static bool is_daemon;
58 static bool get_version; /* IOC getversion IOCTL supported */
59 static char *prog_name;
60 
61 static void G_GNUC_PRINTF(2, 3) do_log(int loglevel, const char *format, ...)
62 {
63     va_list ap;
64 
65     va_start(ap, format);
66     if (is_daemon) {
67         vsyslog(LOG_CRIT, format, ap);
68     } else {
69         vfprintf(stderr, format, ap);
70     }
71     va_end(ap);
72 }
73 
74 static void do_perror(const char *string)
75 {
76     if (is_daemon) {
77         syslog(LOG_CRIT, "%s:%s", string, strerror(errno));
78     } else {
79         fprintf(stderr, "%s:%s\n", string, strerror(errno));
80     }
81 }
82 
83 static int init_capabilities(void)
84 {
85     /* helper needs following capabilities only */
86     int cap_list[] = {
87         CAP_CHOWN,
88         CAP_DAC_OVERRIDE,
89         CAP_FOWNER,
90         CAP_FSETID,
91         CAP_SETGID,
92         CAP_MKNOD,
93         CAP_SETUID,
94     };
95     int i;
96 
97     capng_clear(CAPNG_SELECT_BOTH);
98     for (i = 0; i < ARRAY_SIZE(cap_list); i++) {
99         if (capng_update(CAPNG_ADD, CAPNG_EFFECTIVE | CAPNG_PERMITTED,
100                          cap_list[i]) < 0) {
101             do_perror("capng_update");
102             return -1;
103         }
104     }
105     if (capng_apply(CAPNG_SELECT_BOTH) < 0) {
106         do_perror("capng_apply");
107         return -1;
108     }
109 
110     /* Prepare effective set for setugid.  */
111     for (i = 0; i < ARRAY_SIZE(cap_list); i++) {
112         if (cap_list[i] == CAP_DAC_OVERRIDE) {
113             continue;
114         }
115 
116         if (capng_update(CAPNG_DROP, CAPNG_EFFECTIVE,
117                          cap_list[i]) < 0) {
118             do_perror("capng_update");
119             return -1;
120         }
121     }
122     return 0;
123 }
124 
125 static int socket_read(int sockfd, void *buff, ssize_t size)
126 {
127     ssize_t retval, total = 0;
128 
129     while (size) {
130         retval = read(sockfd, buff, size);
131         if (retval == 0) {
132             return -EIO;
133         }
134         if (retval < 0) {
135             if (errno == EINTR) {
136                 continue;
137             }
138             return -errno;
139         }
140         size -= retval;
141         buff += retval;
142         total += retval;
143     }
144     return total;
145 }
146 
147 static int socket_write(int sockfd, void *buff, ssize_t size)
148 {
149     ssize_t retval, total = 0;
150 
151     while (size) {
152         retval = write(sockfd, buff, size);
153         if (retval < 0) {
154             if (errno == EINTR) {
155                 continue;
156             }
157             return -errno;
158         }
159         size -= retval;
160         buff += retval;
161         total += retval;
162     }
163     return total;
164 }
165 
166 static int read_request(int sockfd, struct iovec *iovec, ProxyHeader *header)
167 {
168     int retval;
169 
170     /*
171      * read the request header.
172      */
173     iovec->iov_len = 0;
174     retval = socket_read(sockfd, iovec->iov_base, PROXY_HDR_SZ);
175     if (retval < 0) {
176         return retval;
177     }
178     iovec->iov_len = PROXY_HDR_SZ;
179     retval = proxy_unmarshal(iovec, 0, "dd", &header->type, &header->size);
180     if (retval < 0) {
181         return retval;
182     }
183     /*
184      * We can't process message.size > PROXY_MAX_IO_SZ.
185      * Treat it as fatal error
186      */
187     if (header->size > PROXY_MAX_IO_SZ) {
188         return -ENOBUFS;
189     }
190     retval = socket_read(sockfd, iovec->iov_base + PROXY_HDR_SZ, header->size);
191     if (retval < 0) {
192         return retval;
193     }
194     iovec->iov_len += header->size;
195     return 0;
196 }
197 
198 static int send_fd(int sockfd, int fd)
199 {
200     struct msghdr msg;
201     struct iovec iov;
202     int retval, data;
203     struct cmsghdr *cmsg;
204     union MsgControl msg_control;
205 
206     iov.iov_base = &data;
207     iov.iov_len = sizeof(data);
208 
209     memset(&msg, 0, sizeof(msg));
210     msg.msg_iov = &iov;
211     msg.msg_iovlen = 1;
212     /* No ancillary data on error */
213     if (fd < 0) {
214         /* fd is really negative errno if the request failed  */
215         data = fd;
216     } else {
217         data = V9FS_FD_VALID;
218         msg.msg_control = &msg_control;
219         msg.msg_controllen = sizeof(msg_control);
220 
221         cmsg = &msg_control.cmsg;
222         cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
223         cmsg->cmsg_level = SOL_SOCKET;
224         cmsg->cmsg_type = SCM_RIGHTS;
225         memcpy(CMSG_DATA(cmsg), &fd, sizeof(fd));
226     }
227 
228     do {
229         retval = sendmsg(sockfd, &msg, 0);
230     } while (retval < 0 && errno == EINTR);
231     if (fd >= 0) {
232         close(fd);
233     }
234     if (retval < 0) {
235         return retval;
236     }
237     return 0;
238 }
239 
240 static int send_status(int sockfd, struct iovec *iovec, int status)
241 {
242     ProxyHeader header;
243     int retval, msg_size;
244 
245     if (status < 0) {
246         header.type = T_ERROR;
247     } else {
248         header.type = T_SUCCESS;
249     }
250     header.size = sizeof(status);
251     /*
252      * marshal the return status. We don't check error.
253      * because we are sure we have enough space for the status
254      */
255     msg_size = proxy_marshal(iovec, 0, "ddd", header.type,
256                              header.size, status);
257     if (msg_size < 0) {
258         return msg_size;
259     }
260     retval = socket_write(sockfd, iovec->iov_base, msg_size);
261     if (retval < 0) {
262         return retval;
263     }
264     return 0;
265 }
266 
267 /*
268  * from man 7 capabilities, section
269  * Effect of User ID Changes on Capabilities:
270  * If the effective user ID is changed from nonzero to 0, then the permitted
271  * set is copied to the effective set.  If the effective user ID is changed
272  * from 0 to nonzero, then all capabilities are are cleared from the effective
273  * set.
274  *
275  * The setfsuid/setfsgid man pages warn that changing the effective user ID may
276  * expose the program to unwanted signals, but this is not true anymore: for an
277  * unprivileged (without CAP_KILL) program to send a signal, the real or
278  * effective user ID of the sending process must equal the real or saved user
279  * ID of the target process.  Even when dropping privileges, it is enough to
280  * keep the saved UID to a "privileged" value and virtfs-proxy-helper won't
281  * be exposed to signals.  So just use setresuid/setresgid.
282  */
283 static int setugid(int uid, int gid, int *suid, int *sgid)
284 {
285     int retval;
286 
287     *suid = geteuid();
288     *sgid = getegid();
289 
290     if (setresgid(-1, gid, *sgid) == -1) {
291         return -errno;
292     }
293 
294     if (setresuid(-1, uid, *suid) == -1) {
295         retval = -errno;
296         goto err_sgid;
297     }
298 
299     if (uid == 0 && gid == 0) {
300         /* Linux has already copied the permitted set to the effective set.  */
301         return 0;
302     }
303 
304     /*
305      * All capabilities have been cleared from the effective set.  However
306      * we still need DAC_OVERRIDE because we don't change supplementary
307      * group ids, and hence may be subject to DAC rules.  init_capabilities
308      * left the set of capabilities that we want in libcap-ng's state.
309      */
310     if (capng_apply(CAPNG_SELECT_CAPS) < 0) {
311         retval = -errno;
312         do_perror("capng_apply");
313         goto err_suid;
314     }
315     return 0;
316 
317 err_suid:
318     if (setresuid(-1, *suid, *suid) == -1) {
319         abort();
320     }
321 err_sgid:
322     if (setresgid(-1, *sgid, *sgid) == -1) {
323         abort();
324     }
325     return retval;
326 }
327 
328 /*
329  * This is used to reset the ugid back with the saved values
330  * There is nothing much we can do checking error values here.
331  */
332 static void resetugid(int suid, int sgid)
333 {
334     if (setresgid(-1, sgid, sgid) == -1) {
335         abort();
336     }
337     if (setresuid(-1, suid, suid) == -1) {
338         abort();
339     }
340 }
341 
342 /*
343  * Open regular file or directory. Attempts to open any special file are
344  * rejected.
345  *
346  * returns file descriptor or -1 on error
347  */
348 static int open_regular(const char *pathname, int flags, mode_t mode)
349 {
350     int fd;
351 
352     fd = open(pathname, flags, mode);
353     if (fd < 0) {
354         return fd;
355     }
356 
357     if (close_if_special_file(fd) < 0) {
358         return -1;
359     }
360 
361     return fd;
362 }
363 
364 /*
365  * send response in two parts
366  * 1) ProxyHeader
367  * 2) Response or error status
368  * This function should be called with marshaled response
369  * send_response constructs header part and error part only.
370  * send response sends {ProxyHeader,Response} if the request was success
371  * otherwise sends {ProxyHeader,error status}
372  */
373 static int send_response(int sock, struct iovec *iovec, int size)
374 {
375     int retval;
376     ProxyHeader header;
377 
378     /*
379      * If response size exceeds available iovec->iov_len,
380      * we return ENOBUFS
381      */
382     if (size > PROXY_MAX_IO_SZ) {
383         size = -ENOBUFS;
384     }
385 
386     if (size < 0) {
387         /*
388          * In case of error we would not have got the error encoded
389          * already so encode the error here.
390          */
391         header.type = T_ERROR;
392         header.size = sizeof(size);
393         proxy_marshal(iovec, PROXY_HDR_SZ, "d", size);
394     } else {
395         header.type = T_SUCCESS;
396         header.size = size;
397     }
398     proxy_marshal(iovec, 0, "dd", header.type, header.size);
399     retval = socket_write(sock, iovec->iov_base, header.size + PROXY_HDR_SZ);
400     if (retval < 0) {
401         return retval;
402     }
403     return 0;
404 }
405 
406 /*
407  * gets generation number
408  * returns -errno on failure and sizeof(generation number) on success
409  */
410 static int do_getversion(struct iovec *iovec, struct iovec *out_iovec)
411 {
412     uint64_t version;
413     int retval = -ENOTTY;
414 #ifdef FS_IOC_GETVERSION
415     int fd;
416     V9fsString path;
417 #endif
418 
419 
420     /* no need to issue ioctl */
421     if (!get_version) {
422         version = 0;
423         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
424         return retval;
425     }
426 #ifdef FS_IOC_GETVERSION
427     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
428     if (retval < 0) {
429         return retval;
430     }
431 
432     fd = open(path.data, O_RDONLY);
433     if (fd < 0) {
434         retval = -errno;
435         goto err_out;
436     }
437     if (ioctl(fd, FS_IOC_GETVERSION, &version) < 0) {
438         retval = -errno;
439     } else {
440         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
441     }
442     close(fd);
443 err_out:
444     v9fs_string_free(&path);
445 #endif
446     return retval;
447 }
448 
449 static int do_getxattr(int type, struct iovec *iovec, struct iovec *out_iovec)
450 {
451     int size = 0, offset, retval;
452     V9fsString path, name, xattr;
453 
454     v9fs_string_init(&xattr);
455     v9fs_string_init(&path);
456     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "ds", &size, &path);
457     if (retval < 0) {
458         return retval;
459     }
460     offset = PROXY_HDR_SZ + retval;
461 
462     if (size) {
463         xattr.data = g_malloc(size);
464         xattr.size = size;
465     }
466     switch (type) {
467     case T_LGETXATTR:
468         v9fs_string_init(&name);
469         retval = proxy_unmarshal(iovec, offset, "s", &name);
470         if (retval > 0) {
471             retval = lgetxattr(path.data, name.data, xattr.data, size);
472             if (retval < 0) {
473                 retval = -errno;
474             } else {
475                 xattr.size = retval;
476             }
477         }
478         v9fs_string_free(&name);
479         break;
480     case T_LLISTXATTR:
481         retval = llistxattr(path.data, xattr.data, size);
482         if (retval < 0) {
483             retval = -errno;
484         } else {
485             xattr.size = retval;
486         }
487         break;
488     }
489     if (retval < 0) {
490         goto err_out;
491     }
492 
493     if (!size) {
494         proxy_marshal(out_iovec, PROXY_HDR_SZ, "d", retval);
495         retval = sizeof(retval);
496     } else {
497         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &xattr);
498     }
499 err_out:
500     v9fs_string_free(&xattr);
501     v9fs_string_free(&path);
502     return retval;
503 }
504 
505 static void stat_to_prstat(ProxyStat *pr_stat, struct stat *stat)
506 {
507     memset(pr_stat, 0, sizeof(*pr_stat));
508     pr_stat->st_dev = stat->st_dev;
509     pr_stat->st_ino = stat->st_ino;
510     pr_stat->st_nlink = stat->st_nlink;
511     pr_stat->st_mode = stat->st_mode;
512     pr_stat->st_uid = stat->st_uid;
513     pr_stat->st_gid = stat->st_gid;
514     pr_stat->st_rdev = stat->st_rdev;
515     pr_stat->st_size = stat->st_size;
516     pr_stat->st_blksize = stat->st_blksize;
517     pr_stat->st_blocks = stat->st_blocks;
518     pr_stat->st_atim_sec = stat->st_atim.tv_sec;
519     pr_stat->st_atim_nsec = stat->st_atim.tv_nsec;
520     pr_stat->st_mtim_sec = stat->st_mtim.tv_sec;
521     pr_stat->st_mtim_nsec = stat->st_mtim.tv_nsec;
522     pr_stat->st_ctim_sec = stat->st_ctim.tv_sec;
523     pr_stat->st_ctim_nsec = stat->st_ctim.tv_nsec;
524 }
525 
526 static void statfs_to_prstatfs(ProxyStatFS *pr_stfs, struct statfs *stfs)
527 {
528     memset(pr_stfs, 0, sizeof(*pr_stfs));
529     pr_stfs->f_type = stfs->f_type;
530     pr_stfs->f_bsize = stfs->f_bsize;
531     pr_stfs->f_blocks = stfs->f_blocks;
532     pr_stfs->f_bfree = stfs->f_bfree;
533     pr_stfs->f_bavail = stfs->f_bavail;
534     pr_stfs->f_files = stfs->f_files;
535     pr_stfs->f_ffree = stfs->f_ffree;
536     pr_stfs->f_fsid[0] = stfs->f_fsid.__val[0];
537     pr_stfs->f_fsid[1] = stfs->f_fsid.__val[1];
538     pr_stfs->f_namelen = stfs->f_namelen;
539     pr_stfs->f_frsize = stfs->f_frsize;
540 }
541 
542 /*
543  * Gets stat/statfs information and packs in out_iovec structure
544  * on success returns number of bytes packed in out_iovec structure
545  * otherwise returns -errno
546  */
547 static int do_stat(int type, struct iovec *iovec, struct iovec *out_iovec)
548 {
549     int retval;
550     V9fsString path;
551     ProxyStat pr_stat;
552     ProxyStatFS pr_stfs;
553     struct stat st_buf;
554     struct statfs stfs_buf;
555 
556     v9fs_string_init(&path);
557     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
558     if (retval < 0) {
559         return retval;
560     }
561 
562     switch (type) {
563     case T_LSTAT:
564         retval = lstat(path.data, &st_buf);
565         if (retval < 0) {
566             retval = -errno;
567         } else {
568             stat_to_prstat(&pr_stat, &st_buf);
569             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
570                                    "qqqdddqqqqqqqqqq", pr_stat.st_dev,
571                                    pr_stat.st_ino, pr_stat.st_nlink,
572                                    pr_stat.st_mode, pr_stat.st_uid,
573                                    pr_stat.st_gid, pr_stat.st_rdev,
574                                    pr_stat.st_size, pr_stat.st_blksize,
575                                    pr_stat.st_blocks,
576                                    pr_stat.st_atim_sec, pr_stat.st_atim_nsec,
577                                    pr_stat.st_mtim_sec, pr_stat.st_mtim_nsec,
578                                    pr_stat.st_ctim_sec, pr_stat.st_ctim_nsec);
579         }
580         break;
581     case T_STATFS:
582         retval = statfs(path.data, &stfs_buf);
583         if (retval < 0) {
584             retval = -errno;
585         } else {
586             statfs_to_prstatfs(&pr_stfs, &stfs_buf);
587             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
588                                    "qqqqqqqqqqq", pr_stfs.f_type,
589                                    pr_stfs.f_bsize, pr_stfs.f_blocks,
590                                    pr_stfs.f_bfree, pr_stfs.f_bavail,
591                                    pr_stfs.f_files, pr_stfs.f_ffree,
592                                    pr_stfs.f_fsid[0], pr_stfs.f_fsid[1],
593                                    pr_stfs.f_namelen, pr_stfs.f_frsize);
594         }
595         break;
596     }
597     v9fs_string_free(&path);
598     return retval;
599 }
600 
601 static int do_readlink(struct iovec *iovec, struct iovec *out_iovec)
602 {
603     char *buffer;
604     int size, retval;
605     V9fsString target, path;
606 
607     v9fs_string_init(&path);
608     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &size);
609     if (retval < 0) {
610         v9fs_string_free(&path);
611         return retval;
612     }
613     buffer = g_malloc(size);
614     v9fs_string_init(&target);
615     retval = readlink(path.data, buffer, size - 1);
616     if (retval > 0) {
617         buffer[retval] = '\0';
618         v9fs_string_sprintf(&target, "%s", buffer);
619         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &target);
620     } else {
621         retval = -errno;
622     }
623     g_free(buffer);
624     v9fs_string_free(&target);
625     v9fs_string_free(&path);
626     return retval;
627 }
628 
629 /*
630  * create other filesystem objects and send 0 on success
631  * return -errno on error
632  */
633 static int do_create_others(int type, struct iovec *iovec)
634 {
635     dev_t rdev;
636     int retval = 0;
637     int offset = PROXY_HDR_SZ;
638     V9fsString oldpath, path;
639     int mode, uid, gid, cur_uid, cur_gid;
640 
641     v9fs_string_init(&path);
642     v9fs_string_init(&oldpath);
643 
644     retval = proxy_unmarshal(iovec, offset, "dd", &uid, &gid);
645     if (retval < 0) {
646         return retval;
647     }
648     offset += retval;
649     retval = setugid(uid, gid, &cur_uid, &cur_gid);
650     if (retval < 0) {
651         goto unmarshal_err_out;
652     }
653     switch (type) {
654     case T_MKNOD:
655         retval = proxy_unmarshal(iovec, offset, "sdq", &path, &mode, &rdev);
656         if (retval < 0) {
657             goto err_out;
658         }
659         retval = mknod(path.data, mode, rdev);
660         break;
661     case T_MKDIR:
662         retval = proxy_unmarshal(iovec, offset, "sd", &path, &mode);
663         if (retval < 0) {
664             goto err_out;
665         }
666         retval = g_mkdir(path.data, mode);
667         break;
668     case T_SYMLINK:
669         retval = proxy_unmarshal(iovec, offset, "ss", &oldpath, &path);
670         if (retval < 0) {
671             goto err_out;
672         }
673         retval = symlink(oldpath.data, path.data);
674         break;
675     }
676     if (retval < 0) {
677         retval = -errno;
678     }
679 
680 err_out:
681     resetugid(cur_uid, cur_gid);
682 unmarshal_err_out:
683     v9fs_string_free(&path);
684     v9fs_string_free(&oldpath);
685     return retval;
686 }
687 
688 /*
689  * create a file and send fd on success
690  * return -errno on error
691  */
692 static int do_create(struct iovec *iovec)
693 {
694     int ret;
695     V9fsString path;
696     int flags, mode, uid, gid, cur_uid, cur_gid;
697 
698     v9fs_string_init(&path);
699     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sdddd",
700                           &path, &flags, &mode, &uid, &gid);
701     if (ret < 0) {
702         goto unmarshal_err_out;
703     }
704     ret = setugid(uid, gid, &cur_uid, &cur_gid);
705     if (ret < 0) {
706         goto unmarshal_err_out;
707     }
708     ret = open_regular(path.data, flags, mode);
709     if (ret < 0) {
710         ret = -errno;
711     }
712 
713     resetugid(cur_uid, cur_gid);
714 unmarshal_err_out:
715     v9fs_string_free(&path);
716     return ret;
717 }
718 
719 /*
720  * open a file and send fd on success
721  * return -errno on error
722  */
723 static int do_open(struct iovec *iovec)
724 {
725     int flags, ret;
726     V9fsString path;
727 
728     v9fs_string_init(&path);
729     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &flags);
730     if (ret < 0) {
731         goto err_out;
732     }
733     ret = open_regular(path.data, flags, 0);
734     if (ret < 0) {
735         ret = -errno;
736     }
737 err_out:
738     v9fs_string_free(&path);
739     return ret;
740 }
741 
742 /* create unix domain socket and return the descriptor */
743 static int proxy_socket(const char *path, uid_t uid, gid_t gid)
744 {
745     int sock, client;
746     struct sockaddr_un proxy, qemu;
747     socklen_t size;
748 
749     /* requested socket already exists, refuse to start */
750     if (!access(path, F_OK)) {
751         do_log(LOG_CRIT, "socket already exists\n");
752         return -1;
753     }
754 
755     if (strlen(path) >= sizeof(proxy.sun_path)) {
756         do_log(LOG_CRIT, "UNIX domain socket path exceeds %zu characters\n",
757                sizeof(proxy.sun_path));
758         return -1;
759     }
760 
761     sock = socket(AF_UNIX, SOCK_STREAM, 0);
762     if (sock < 0) {
763         do_perror("socket");
764         return -1;
765     }
766 
767     /* mask other part of mode bits */
768     umask(7);
769 
770     proxy.sun_family = AF_UNIX;
771     strcpy(proxy.sun_path, path);
772     if (bind(sock, (struct sockaddr *)&proxy,
773             sizeof(struct sockaddr_un)) < 0) {
774         do_perror("bind");
775         goto error;
776     }
777     if (chown(proxy.sun_path, uid, gid) < 0) {
778         do_perror("chown");
779         goto error;
780     }
781     if (listen(sock, 1) < 0) {
782         do_perror("listen");
783         goto error;
784     }
785 
786     size = sizeof(qemu);
787     client = accept(sock, (struct sockaddr *)&qemu, &size);
788     if (client < 0) {
789         do_perror("accept");
790         goto error;
791     }
792     close(sock);
793     return client;
794 
795 error:
796     close(sock);
797     return -1;
798 }
799 
800 static void usage(void)
801 {
802     fprintf(stderr, "usage: %s\n"
803             " -p|--path <path> 9p path to export\n"
804             " {-f|--fd <socket-descriptor>} socket file descriptor to be used\n"
805             " {-s|--socket <socketname> socket file used for communication\n"
806             " \t-u|--uid <uid> -g|--gid <gid>} - uid:gid combination to give "
807             " access to this socket\n"
808             " \tNote: -s & -f can not be used together\n"
809             " [-n|--nodaemon] Run as a normal program\n",
810             prog_name);
811 }
812 
813 static int process_reply(int sock, int type,
814                          struct iovec *out_iovec, int retval)
815 {
816     switch (type) {
817     case T_OPEN:
818     case T_CREATE:
819         if (send_fd(sock, retval) < 0) {
820             return -1;
821         }
822         break;
823     case T_MKNOD:
824     case T_MKDIR:
825     case T_SYMLINK:
826     case T_LINK:
827     case T_CHMOD:
828     case T_CHOWN:
829     case T_TRUNCATE:
830     case T_UTIME:
831     case T_RENAME:
832     case T_REMOVE:
833     case T_LSETXATTR:
834     case T_LREMOVEXATTR:
835         if (send_status(sock, out_iovec, retval) < 0) {
836             return -1;
837         }
838         break;
839     case T_LSTAT:
840     case T_STATFS:
841     case T_READLINK:
842     case T_LGETXATTR:
843     case T_LLISTXATTR:
844     case T_GETVERSION:
845         if (send_response(sock, out_iovec, retval) < 0) {
846             return -1;
847         }
848         break;
849     default:
850         return -1;
851         break;
852     }
853     return 0;
854 }
855 
856 static int process_requests(int sock)
857 {
858     int flags;
859     int size = 0;
860     int retval = 0;
861     uint64_t offset;
862     ProxyHeader header;
863     int mode, uid, gid;
864     V9fsString name, value;
865     struct timespec spec[2];
866     V9fsString oldpath, path;
867     struct iovec in_iovec, out_iovec;
868 
869     in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
870     in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
871     out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
872     out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
873 
874     while (1) {
875         /*
876          * initialize the header type, so that we send
877          * response to proper request type.
878          */
879         header.type = 0;
880         retval = read_request(sock, &in_iovec, &header);
881         if (retval < 0) {
882             goto err_out;
883         }
884 
885         switch (header.type) {
886         case T_OPEN:
887             retval = do_open(&in_iovec);
888             break;
889         case T_CREATE:
890             retval = do_create(&in_iovec);
891             break;
892         case T_MKNOD:
893         case T_MKDIR:
894         case T_SYMLINK:
895             retval = do_create_others(header.type, &in_iovec);
896             break;
897         case T_LINK:
898             v9fs_string_init(&path);
899             v9fs_string_init(&oldpath);
900             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
901                                      "ss", &oldpath, &path);
902             if (retval > 0) {
903                 retval = link(oldpath.data, path.data);
904                 if (retval < 0) {
905                     retval = -errno;
906                 }
907             }
908             v9fs_string_free(&oldpath);
909             v9fs_string_free(&path);
910             break;
911         case T_LSTAT:
912         case T_STATFS:
913             retval = do_stat(header.type, &in_iovec, &out_iovec);
914             break;
915         case T_READLINK:
916             retval = do_readlink(&in_iovec, &out_iovec);
917             break;
918         case T_CHMOD:
919             v9fs_string_init(&path);
920             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
921                                      "sd", &path, &mode);
922             if (retval > 0) {
923                 retval = chmod(path.data, mode);
924                 if (retval < 0) {
925                     retval = -errno;
926                 }
927             }
928             v9fs_string_free(&path);
929             break;
930         case T_CHOWN:
931             v9fs_string_init(&path);
932             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sdd", &path,
933                                      &uid, &gid);
934             if (retval > 0) {
935                 retval = lchown(path.data, uid, gid);
936                 if (retval < 0) {
937                     retval = -errno;
938                 }
939             }
940             v9fs_string_free(&path);
941             break;
942         case T_TRUNCATE:
943             v9fs_string_init(&path);
944             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sq",
945                                      &path, &offset);
946             if (retval > 0) {
947                 retval = truncate(path.data, offset);
948                 if (retval < 0) {
949                     retval = -errno;
950                 }
951             }
952             v9fs_string_free(&path);
953             break;
954         case T_UTIME:
955             v9fs_string_init(&path);
956             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sqqqq", &path,
957                                      &spec[0].tv_sec, &spec[0].tv_nsec,
958                                      &spec[1].tv_sec, &spec[1].tv_nsec);
959             if (retval > 0) {
960                 retval = utimensat(AT_FDCWD, path.data, spec,
961                                    AT_SYMLINK_NOFOLLOW);
962                 if (retval < 0) {
963                     retval = -errno;
964                 }
965             }
966             v9fs_string_free(&path);
967             break;
968         case T_RENAME:
969             v9fs_string_init(&path);
970             v9fs_string_init(&oldpath);
971             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
972                                      "ss", &oldpath, &path);
973             if (retval > 0) {
974                 retval = rename(oldpath.data, path.data);
975                 if (retval < 0) {
976                     retval = -errno;
977                 }
978             }
979             v9fs_string_free(&oldpath);
980             v9fs_string_free(&path);
981             break;
982         case T_REMOVE:
983             v9fs_string_init(&path);
984             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "s", &path);
985             if (retval > 0) {
986                 retval = remove(path.data);
987                 if (retval < 0) {
988                     retval = -errno;
989                 }
990             }
991             v9fs_string_free(&path);
992             break;
993         case T_LGETXATTR:
994         case T_LLISTXATTR:
995             retval = do_getxattr(header.type, &in_iovec, &out_iovec);
996             break;
997         case T_LSETXATTR:
998             v9fs_string_init(&path);
999             v9fs_string_init(&name);
1000             v9fs_string_init(&value);
1001             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sssdd", &path,
1002                                      &name, &value, &size, &flags);
1003             if (retval > 0) {
1004                 retval = lsetxattr(path.data,
1005                                    name.data, value.data, size, flags);
1006                 if (retval < 0) {
1007                     retval = -errno;
1008                 }
1009             }
1010             v9fs_string_free(&path);
1011             v9fs_string_free(&name);
1012             v9fs_string_free(&value);
1013             break;
1014         case T_LREMOVEXATTR:
1015             v9fs_string_init(&path);
1016             v9fs_string_init(&name);
1017             retval = proxy_unmarshal(&in_iovec,
1018                                      PROXY_HDR_SZ, "ss", &path, &name);
1019             if (retval > 0) {
1020                 retval = lremovexattr(path.data, name.data);
1021                 if (retval < 0) {
1022                     retval = -errno;
1023                 }
1024             }
1025             v9fs_string_free(&path);
1026             v9fs_string_free(&name);
1027             break;
1028         case T_GETVERSION:
1029             retval = do_getversion(&in_iovec, &out_iovec);
1030             break;
1031         default:
1032             goto err_out;
1033             break;
1034         }
1035 
1036         if (process_reply(sock, header.type, &out_iovec, retval) < 0) {
1037             goto err_out;
1038         }
1039     }
1040 err_out:
1041     g_free(in_iovec.iov_base);
1042     g_free(out_iovec.iov_base);
1043     return -1;
1044 }
1045 
1046 int main(int argc, char **argv)
1047 {
1048     int sock;
1049     uid_t own_u;
1050     gid_t own_g;
1051     char *rpath = NULL;
1052     char *sock_name = NULL;
1053     struct stat stbuf;
1054     int c, option_index;
1055 #ifdef FS_IOC_GETVERSION
1056     int retval;
1057     struct statfs st_fs;
1058 #endif
1059 
1060     prog_name = g_path_get_basename(argv[0]);
1061 
1062     is_daemon = true;
1063     sock = -1;
1064     own_u = own_g = -1;
1065     while (1) {
1066         option_index = 0;
1067         c = getopt_long(argc, argv, "p:nh?f:s:u:g:", helper_opts,
1068                         &option_index);
1069         if (c == -1) {
1070             break;
1071         }
1072         switch (c) {
1073         case 'p':
1074             rpath = g_strdup(optarg);
1075             break;
1076         case 'n':
1077             is_daemon = false;
1078             break;
1079         case 'f':
1080             sock = atoi(optarg);
1081             break;
1082         case 's':
1083             sock_name = g_strdup(optarg);
1084             break;
1085         case 'u':
1086             own_u = atoi(optarg);
1087             break;
1088         case 'g':
1089             own_g = atoi(optarg);
1090             break;
1091         case '?':
1092         case 'h':
1093         default:
1094             usage();
1095             exit(EXIT_FAILURE);
1096         }
1097     }
1098 
1099     /* Parameter validation */
1100     if ((sock_name == NULL && sock == -1) || rpath == NULL) {
1101         fprintf(stderr, "socket, socket descriptor or path not specified\n");
1102         usage();
1103         return -1;
1104     }
1105 
1106     if (sock_name && sock != -1) {
1107         fprintf(stderr, "both named socket and socket descriptor specified\n");
1108         usage();
1109         exit(EXIT_FAILURE);
1110     }
1111 
1112     if (sock_name && (own_u == -1 || own_g == -1)) {
1113         fprintf(stderr, "owner uid:gid not specified, ");
1114         fprintf(stderr,
1115                 "owner uid:gid specifies who can access the socket file\n");
1116         usage();
1117         exit(EXIT_FAILURE);
1118     }
1119 
1120     if (lstat(rpath, &stbuf) < 0) {
1121         fprintf(stderr, "invalid path \"%s\" specified, %s\n",
1122                 rpath, strerror(errno));
1123         exit(EXIT_FAILURE);
1124     }
1125 
1126     if (!S_ISDIR(stbuf.st_mode)) {
1127         fprintf(stderr, "specified path \"%s\" is not directory\n", rpath);
1128         exit(EXIT_FAILURE);
1129     }
1130 
1131     if (is_daemon) {
1132         if (daemon(0, 0) < 0) {
1133             fprintf(stderr, "daemon call failed\n");
1134             exit(EXIT_FAILURE);
1135         }
1136         openlog(PROGNAME, LOG_PID, LOG_DAEMON);
1137     }
1138 
1139     do_log(LOG_INFO, "Started\n");
1140     if (sock_name) {
1141         sock = proxy_socket(sock_name, own_u, own_g);
1142         if (sock < 0) {
1143             goto error;
1144         }
1145     }
1146 
1147     if (chroot(rpath) < 0) {
1148         do_perror("chroot");
1149         goto error;
1150     }
1151     if (chdir("/") < 0) {
1152         do_perror("chdir");
1153         goto error;
1154     }
1155 
1156     get_version = false;
1157 #ifdef FS_IOC_GETVERSION
1158     /* check whether underlying FS support IOC_GETVERSION */
1159     retval = statfs("/", &st_fs);
1160     if (!retval) {
1161         switch (st_fs.f_type) {
1162         case EXT2_SUPER_MAGIC:
1163         case BTRFS_SUPER_MAGIC:
1164         case REISERFS_SUPER_MAGIC:
1165         case XFS_SUPER_MAGIC:
1166             get_version = true;
1167             break;
1168         }
1169     }
1170 #endif
1171 
1172     umask(0);
1173     if (init_capabilities() < 0) {
1174         goto error;
1175     }
1176 
1177     process_requests(sock);
1178 error:
1179     g_free(rpath);
1180     g_free(sock_name);
1181     do_log(LOG_INFO, "Done\n");
1182     closelog();
1183     return 0;
1184 }
1185